{"id":"raon-1-mini-krafton","slug":"raon-1-mini-krafton","name":"Raon-1 mini (preview)","orgId":"krafton","family":"raon-1-mini-krafton","releaseDate":null,"availability":"announced","openWeights":true,"license":null,"paramsTotal":null,"paramsSource":null,"contextWindow":null,"knowledgeCutoff":null,"modelCardUrl":"https://x.com/Krafton_AI/status/2112772568776339709","hfRepo":null,"modalities":["text"],"aliases":["Raon-1 mini (preview)","raon-1 mini","raon-1-mini","raon 1 mini","raon-1 mini preview","raon-1-mini-preview","krafton/raon-1-mini-preview","krafton raon-1 mini"],"aliasNote":null,"origin":"tracked","tracked":{"trackedSince":"2026-10-21","announced":"2026-10-21","blurb":"KRAFTON AI's first general language model, a sparse mixture-of-experts model trained from scratch, released as a preview. It mixes Raon Delta Attention (linear attention) with gated MLA layers, natively handles 262,144 tokens and is tuned for Korean conversation and agentic tasks. Weights are on Hugging Face under Apache 2.0, behind an automatically approved access request.","api":"waitlist","openWeights":"yes","params":{"text":"28.8B total, 4.3B active (marketed as 30B-A4B)","source":"KRAFTON's Hugging Face model card"},"facts":[{"label":"Status","value":"Preview; KRAFTON says specs and results may change before the final release"},{"label":"Architecture","value":"MoE, 25 layers: 18 Raon Delta Attention (linear) and 7 gated MLA; 6 of 128 routed experts plus 1 shared active per token; no positional encoding"},{"label":"Context","value":"262,144 tokens (native)"},{"label":"Training","value":"Trained from scratch by KRAFTON AI; languages: Korean, English, Japanese, Chinese, Spanish, French and Thai"},{"label":"Weights","value":"Apache 2.0 on Hugging Face (gated, approval is automatic), in MXFP8 and NVFP4, plus GGUF, MLX and a DFlash2 speculative-decoding drafter"},{"label":"Hosted access","value":"Free playground for a limited time; hosted API in private beta by email request, under separate terms (not Apache 2.0)"}],"claims":["Lab claim: matches Muse Glimmer on coding and agentic tasks while being faster and more memory-efficient. The model card's own table is mixed against Muse Glimmer 30B (for example SWE-bench Verified 70.8 vs 72.0, Terminal-Bench Lite 47.0 vs 70.7, BrowseComp Plus 55.7 vs 52.8).","Lab claim: the DFlash2 drafter gives up to 2.18× higher output throughput than autoregressive decoding on a single NVIDIA B300."],"sources":[{"kind":"lab","title":"KRAFTON AI announcement on X","url":"https://x.com/Krafton_AI/status/2112772568776339709","date":"2026-10-21"},{"kind":"repo","title":"Raon-1-mini-preview on Hugging Face","url":"https://huggingface.co/KRAFTON/Raon-1-mini-preview","date":"2026-10-21"},{"kind":"lab","title":"Raon-1 mini playground","url":"https://raon-mini.krafton.ai/en","date":"2026-10-21"}],"firstBenchmark":null},"organization":{"id":"krafton","name":"KRAFTON AI","slug":"krafton","country":null,"url":"https://raon-mini.krafton.ai/en","color":"#11110F","mark":"K"},"price":null,"answer":{"word":"Pending benchmarks.","scope":"Announced Oct 21, 2026; no independent results yet.","lead":"As of Oct 22, 2026, Raon-1 mini (preview) is pending benchmarks on isitsota: announced Oct 21, 2026, with no independent results on the boards we track yet.","confidence":"low"},"status":"PENDING_BENCHMARKS","verdicts":[]}